REASONS ARE CONSIDERATIONS THAT FIGURE in reasoning. When I believe or act for a reason, my reason plays a role in the reasoning through which I arrive at my belief or action. Of course, not all of the reasons for which I believe or act are good reasons; not all explanatory reasons are normative or justifying reasons. (1) If reasons in general are considerations that figure in reasoning, normative reasons are considerations that figure in sound reasoning.
This broad view about the connection between normative reasons and sound reasoning is widely held. (2) Those who hold it often regard it as obvious--boring, even. As Kieran Setiya puts it:
It is a harmlessly illuminating principle that connects two things which surely must be connected: facts being reasons on one side, and the process of ... thinking, inference, deliberation, on the other. (3) Jonathan Way describes the link between reasons and reasoning as "near platitudinous":
Reasons are meant to guide us to act, believe, desire, or otherwise respond. But to be guided by reasons just is to engage in reasoning, broadly construed. So it is hard to see how reasons could fail to be appropriate premises for reasoning towards [phi]-ing. (4) The idea of a connection between normative reasons and sound reasoning does seem platitudinous. However, I think it actually has surprising and far-reaching metanormative implications. The view that reasons are linked to sound reasoning may strike us as "harmlessly illuminating," but only because we tend to assume that soundness is a normative property, in which case the view merely relates one normative phenomenon (reasons) to another (soundness). I shall argue, though, that soundness is also a descriptive phenomenon, one we can pick out with purely descriptive terms, and that the connection between normative reasons and sound reasoning therefore provides the basis for a reductive account of reasons. Like all proposed reductions, this one must confront some version of G. E. Moore's open question argument. I shall argue that a reductive view rooted in the idea that reasons figure in sound reasoning is well equipped to meet the open question challenge head on.
Following both Setiya and Way, I shall take as my starting point the idea that normative reasons are premises in sound reasoning. More specifically, I shall follow Way in supposing that a consideration R is a reason for some response 9 just in case R is a premise in sound reasoning that concludes in [phi]-ing. (5)
As Way has pointed out, the hypothesis that reasons are premises of sound reasoning sheds light on a number of puzzles about reasons. It explains what the various kinds of reasons--reasons for action, reasons for belief, reasons for desire, and so forth--have in common, why we can have normative reasons only for attitudes we can reach through reasoning, and why some of the considerations that count in favor of a response are the wrong kind of reason for that response. (6) The hypothesis also raises a number of questions, the most pressing of which concerns the nature of soundness. Just what is sound reasoning? The short answer is that sound reasoning is good or correct reasoning: to reason soundly is to reason well--to reason as one should. If that is all we can say about soundness, then the hypothesis that normative reasons figure in sound reasoning really will be harmless. I believe we can say more, though. I accept the short answer: sound reasoning is good or correct reasoning, and so soundness is indeed a normative property. However, I think it is also a descriptive property--a property that can be grasped or picked out in purely descriptive terms. To argue for this surprising conclusion, I shall begin with the case of doxastic reasoning, or reasoning about what to believe. I shall then argue that we can understand the soundness of any form of reasoning along the same lines we understand the soundness of doxastic reasoning.
As we teach students in introductory logic courses, deductively sound doxastic reasoning is just deductively valid doxastic reasoning from true premises. Truth is not a normative concept. Nor, plausibly, is validity. An argument or a bit of reasoning is valid just in case there are no possible worlds in which all of the premises are true and the conclusion false. Validity therefore looks more like a modal concept than a normative one. And in that case soundness does as well: we can identify an instance of deductive doxastic reasoning as sound without deploying any normative concepts. (7)
This appearance could be misleading, of course. After all, sound reasoning is supposed to be good or correct reasoning, and there are all sorts of unsettled questions about the relation between valid reasoning and good or correct reasoning. It is widely accepted, for instance, that not all valid inferences are good inferences. Suppose that I believe that p and that I come to believe that if p, then q. If I then infer that q, my inference is clearly valid. Yet it may not be the correct inference: if I have strong independent reasons to think that q is false and relatively weak reasons to believe that p, then surely I ought to give up my belief that p. Surely, in other words, I should reason in accordance with modus tollens here rather than modusponens. Either inference would be valid, but only one of them is correct. Likewise, if I reason from contradictory premises to the conclusion that all unicorns are white, then--on at least some accounts of deductive validity--my reasoning is valid. But it is also clearly problematic. We must therefore be careful before concluding that we can understand sound deductive reasoning in purely logical terms.
Yet in the cases where deductively valid reasoning falls short of good or correct reasoning, adding that the premises are true takes us the rest of the way. Given any conditional, for instance, we can reason soundly in only one direction. Suppose that I correctly believe that if p, then q. If p is true, then modus ponens is the sound inference. If p is false, then modus tollens is the sound inference. Similarly, we need not worry about inferences from contradictory premises, since no such inference can be sound. This suggests that we do not need to settle questions about the relation between valid reasoning and correct reasoning in order to capture the soundness of deductive doxastic reasoning in purely descriptive terms. There are no instances of deductively sound reasoning that fail to be instances of good or correct reasoning, even though there are instances of bad or incorrect deductively valid reasoning.
There is more to doxastic reasoning than deductive syllogisms, however. Most doxastic deliberation consists of inductive or abductive reasoning. Yet it is quite plausible that we can also make sense of the soundness of these other forms of reasoning in purely descriptive terms. As in the deductive case, sound inductive reasoning proceeds from true premises. The connection between the truth of the premises and the truth of the conclusion is much looser when it comes to inductive reasoning, though. In a sound inductive inference, although the truth of the premises does not necessitate the truth of the conclusion, it does probabilize the conclusion or render it more likely. As formal epistemologists are fond of saying, it confirms the conclusion. Of course, it also justifies or counts in favor of our believing the conclusion, and thus it is tempting to conclude that the relation between the premises of a sound inductive inference and the conclusion is a normative one. We need not resist that temptation: it is indeed a normative relation. But it is also a descriptive one--one we can grasp in purely descriptive terms. Although sound inductive reasoning is just correct inductive reasoning, it is also just reasoning in which true premises probabilize or confirm the conclusion. If this is correct, then the soundness of doxastic reasoning is a descriptive property as well as a normative one.
We might want to reserve the honorific "sound" for instances of doxastic reasoning in which one takes into account all of the available considerations that bear on the truth of the proposition about which one is deliberating. Suppose I notice that there is water falling outside my window and then conclude on the basis of that observation that it is raining. Is my reasoning sound? It may depend on what else I know. For instance, if I know that it almost never rains here, that there is not a cloud in the sky, and that the windows above mine are currently being washed, then presumably I would not be reasoning soundly were I to conclude that it is raining. My reasoning would be sound only if I accorded due weight to these other relevant considerations, and were I to do so I would quickly conclude that the water falling outside my window has nothing to do with rain. There appears, then, to be more to sound doxastic reasoning than realizing that some consideration confirms or otherwise supports the truth of some conclusion and then accepting that conclusion on the basis of that realization.
We can still grasp the differences between sound and unsound doxastic reasoning in purely descriptive terms, however. We can capture the full scope of this stronger conception of sound reasoning by taking our original account and adding that the premises of sound reasoning must include all of one's true beliefs that bear one way or the other on the truth of the conclusion. This ideal of soundness is probably never realized, of course. But the various ways in which our doxastic reasoning falls short can be articulated descriptively. Sometimes we draw conclusions that are not confirmed or otherwise supported by our premises. Other times we fail to incorporate relevant considerations in our deliberations. And other times we reason from false premises. So even though soundness with respect to doxastic reasoning comes in degrees, the differences among those...